buildr-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From lacton <>
Subject Re: Changes and backward compatibility
Date Tue, 09 Sep 2008 18:00:40 GMT
On Tue, Sep 9, 2008 at 5:28 AM, Assaf Arkin <> wrote:
> Every release we make is an opportunity to fix all the reported issues
> and bugs.  And to break things that worked before.  Sometimes both at
> the same time, which to enough people translate into "nothing worth
> upgrading for".

I know what you mean.  I have this issue with release 1.3.2.  It has
several enhancement I'd like for my day job, but issue BUILDR-106 is a
blocker, so my team is sticking with

> We need to be careful with what we change, how we change, and how we
> roll out those changes.  Especially changes that "remove" features,
> whether by changing an API, behavior or simply breaking something that
> worked before.
> I'd like to propose the following policy:
> 1. Concentrate major changes along minor/major version boundaries, not
> point releases (i.e. 1.3, 1.4, 1.5, 2.0, etc).

I agree with this idea, but I wonder what it means from an SVN point
of view.  When we add a major change, should we do it on trunk? on a

> 2. Maintain backward compatibility for at least one minor version.
> For example, if we decide to remove a feature and the current release
> is 1.3.4, we retain that feature for one minor version (1.4.x) up
> until the 1.5.0 release.

Fine with me.

> 3. Provide deprecated warnings in documentation, API and when running
> code. The warning should indicate what alternative exists, e.g.
> "Please replace with using(:foo=>true)".

Yes.  That's something that's been done quite well in buildr.

> 4. Maintain specs/tests for both new and deprecated behavior, until
> deprecated behavior has been removed.

To document this, I suggest we add expectations in our specs.


it 'should do a deprecated behavior' do
  Buildr.VERSION.should < '1.5'

That way, when we reach version 1.5, we'll know automatically what
needs to be removed. :-)

> 5. Obviously there will be exceptions to this rule, but having a rule
> we can at least minimize/contain these exceptions and know how to deal
> with them.



View raw message